Visual analysis of differential evolution algorithms

Abstract

In this article a web tool which contributes to the visual analysis of the Differential Evolution (DE) algorithms is presented. The tool provides a graphic interface with 8 views that allows understanding the underlying process of the algorithm. The tool has a library which extracts data from DE algorithms and its main feature is that the functions of the library can be embedded in the code of any DE algorithm to be analyzed. To validate the tool, three DE algorithms: DE/Rand/1/bin, DE/best/1/bin, and JADE and three test functions: Sphere, Rosenbrock, and Rastrigin have been used, which produced a total of 234 different tests, all of them performed successfully. The tool can allow to experts to analyze algorithms, particularly DE algorithms, and it can contribute to improve such algorithms or in generating new strategies that can emerge from the analysis of the extracted information.